Submission doc.: IEEE /0612r0 May 2014 Jiyong Pang, et. al. HuaweiSlide 1 Calibration Procedures towards Integrated System Level Simulation Date:

Submission doc.: IEEE /0612r0 Abstract This presentation discusses calibration procedures contributing to an integrated system level simulation, with emphasis on metrics and methods of each calibration box. The presentation provides preliminary results of instantaneous SINR, PHY SLS and Integrated SLS of HEW scenarios to promote the calibration progress. Slide 2 May 2014 Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0 Background Step-by-step calibration procedures of integrated system level simulation were discussed in [1, 2]. Six companies have validated Long-term SINR calibration [3] for HEW scenarios of residential, enterprise, indoor and outdoor defined in [4]. According to recent offline discussion on evaluation methodology, calibration procedures are recommended to be divided into several boxes: Box 0 - PHY abstraction Box 1 - Long term SINR calibration Box 2 - Instantaneous SINR calibration Box 3 - MAC SLS calibration Box 4 - PHY SLS calibration Box 5 - Integrated SLS calibration Slide 3 May 2014 Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0 Box 0: PHY Abstraction Slide 4 May 2014 Jiyong Pang, et. al. Huawei In [5]~[8], several ESM based PHY abstraction methods were proposed to predict instantaneous PER, e.g. RBIR-CM/BICM, MMIB and capacity based ESM. Considering the workload and difficulty to achieve consensus on a common PHY abstraction in a reasonable time period, and given that a common PHY abstraction method is not prerequisite so long as various PHY abstraction methods are sufficiently calibrated and validated, a generic PHY abstraction box 0 is warranted Objective: Ensure different PHY abstraction methods will provide similar PER prediction. Method: Compare the PER obtained from PHY abstraction methods and reference curves from link level simulation under same channel assumption. Metric: Opt 1: Effective SNR vs. PER (for all channel models) Opt 2: Average SNR vs. PER (for each specific channel model) In [5][6], RBIR was shown to provide a PER prediction with sufficient accuracy for integrated SLS. The calibration of AWGN curves to be used as reference curves in ESM is a good starting point.

Submission doc.: IEEE /0612r0 Box 1: Long Term SINR/SNR Calibration Slide 5 May 2014 Objective Calibrate deployment, large scale fading/shadowing, station association Method No need to realize any channel access mechanism in this step Several options were discussed on how to calculate the OBSS interference Opt 1[9]: interference from all OBSS AP (DL only) Opt 2[9]: interference from one randomly selected STA per OBSS (UL only) Opt 3[9]: interference from one randomly selected STA or AP per OBSS (50/50 DL&UL) Opt 4[2]: average interference over all nodes within each OBSS (i.e., every single interference term is weighted by one over the total number of nodes in that BSS.) All communication links are considered in one calculation. Six companies already achieved alignment with Opt 4 for all HEW scenarios. [3] Metric Separate DL/UL CDF of Long term SINR/SNR of each potential communication link Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0May 2014 Slide 6 Objective Calibrate small scale fading channel of communication links and interference links Method Simple MAC is necessary to decide node to transmit. Opt 1[9]: Dl only; Opt 2[9]: UL only; Opt 3[9]: 50/50 DL&UL; Opt 4: CCA-only mechanism [10] with assumption of all nodes having the same channel access probability, which can be reused in the PHY SLS (i.e., box 4) Opt 5[11]: EDCA need further calibration in MAC Full buffer traffic No need for MIMO calibration with MMSE receiver 1 Tx and 1 Rx are assumed; no antenna gain and cable loss Metric Opt 1: Separate CDF of DL/UL Instantaneous SINR/SNR per tone of all OFDM symbols Offer all details for SINR distribution and can be reused by all effective SINR calculation Opt 2: capacity based Effective SINR There is no consensus on which effective SINR to use for PHY abstraction. Box 2: Instantaneous SINR Calibration Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0 Simple MAC - CCA-Only 1. for drop = 1 : D 2. { 3. Drop STAs/APs and associate STAs with APs according to the deployment requirement defined by the scenario; 4. Assign each node (APs/STAs) an exclusive ordering_ID ∈ [1, N] where N is the total number of all network nodes; 4. for slot = 1 : S 5. { 6. All nodes are not activated at the beginning 7. Generate an N-element random array R(1:N) where R(n) ∈ [1, N]; 8. for n = 1 : N 9. { // Check whether the n-th node with ordering_ID =R(n) could be a transmitter ? 10. The n-th node is just the node R(n) ; 11. if (all BSSs have selected one and only one transmitter) 12. break; 13. elseif (the node R(n) has been activated passively as a receiver) 14. continue; 15. elseif (there is no activated transmitter in the BSS the node R(n) associated with) 16. The node R(n) checks its CCA status based on interference level from transmitter nodes that are already activated 17. if (CCA idle), 18. The node R(n) is activated as a transmitter; 19. if (the node R(n) is a STA), 20. its associated AP is also activated as the receiver; 21. if (the node R(n) is an AP), 22. one STA within the same BSS is randomly activated as the receiver; 23. elseif (CCA busy) 24. continue; 25. else 26. continue; 27. } 28. Instantaneous SINR for STAs (downlink) and APs (uplink) is collected based on current-slot Tx-Rx profile } 4. } 5. Generate CDF of instantaneous SINR collected over multiple drops Jiyong Pang, et. al. HuaweiSlide 7 A fixed CCA threshold for all nodes (e.g., -82dBm) In each drop, assign each node (AP or STA) an exclusive sequence number from 1 to N (where N is the total number of all nodes) Slot-by-slot simulation - At the start of each slot, generate N different random numbers within [1, N] -Select a node one-by-one from the random number array. A transmitter is selected only if the CCA is idle at this node. Only single node-pair transmission is allowed in each BSS -SINR is computed at the active receiving nodes once the transmission node has been selected. May 2014

Submission doc.: IEEE /0612r0 Slide 8 May 2014 Box 2: Instantaneous SNR/SINR Calibration Jiyong Pang, et. al. Huawei These simulation results have been calibrated and validated with some other companies

Submission doc.: IEEE /0612r0 Slide 9 May 2014 Box 2: Instantaneous SNR/SINR Calibration Jiyong Pang, et. al. Huawei These simulation results have been calibrated and validated with some other companies

Submission doc.: IEEE /0612r0May 2014 Slide 10 Objective: Calibrate the behavior of MAC functions Method Use simulation log to verify the protocol implementation of basic MAC features on point-to-point link; Performance test in multiple STA scenarios No need to include PHY features in MAC calibration because the results of other boxes, prior calibrated results can be reused. Refer to our contribution for more details [12] Box 3: MAC SLS Calibration Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0May 2014 Slide 11 Objective Provide accurate modeling of the PHY SLS using PHY abstraction and provide PHY SLS performance baseline Metric: CDF of per non-AP STA throughput Assumption Full buffer traffic; Simple MAC mechanism, same as Box 2; PHY abstraction (such as RBIR); PHY features such as MIMO(e.g. 2x2 ), TxBF, link adaption and receiver algorithm (e.g MMSE) Box 4: PHY SLS Calibration Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0 Box 4 : PHY SLS Calibration Slide 12 May 2014 Jiyong Pang, et. al. Huawei These simulation results have not been validated with other companies PHY Assumptions MIMO2x2 channel B TxBFIdeal CSI feedback, eigen vector beamforming with rank adaption ReceiverMMSE MCS selectionmax goodput ensuring <10% PER PHY-absRBIR Time slot1ms

Submission doc.: IEEE /0612r0May 2014 Slide 13 Objective Calibrate full functions of system level simulation Provide aligned 11ac performance baseline with both PHY and MAC Metric CDF of per non-AP STA throughput Assumption Use full buffer traffic for function tests Use real traffic model, traffic mix and loading from the simulations scenarios document to obtain 11ac performance baseline. Remarks As we need to evaluate HEW techniques in each HEW scenario, a common baseline performance for each scenario is required. Box 5: Integrated SLS Calibration Feature Minimum List MAC CCA Control frame (RTS/CTS/ACK/Block ACK) EDCA Aggregation (A-MPDU in 11ac) Link Adaption Transmission mode (SU-OL, Beamforming,…) selection PHY Beamforming vector MMSE Effective SINR Mapping and PER prediction Energy detection Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0 Box 5: Integrated SLS Calibration Slide 14 May 2014 Jiyong Pang, et. al. Huawei These simulation results have not been validated with other companies MAC Assumptions EDCAon Control frameRTS/CTS/BA/AK TXOPAC_VI 3ms AC_VO 1.5ms AC_BE/AC_BK no limit RTS threshold1000bytes A-MPDUEach MPDU 1500bytes Max 64 MPDU Simulation time3s per drop

Submission doc.: IEEE /0612r0May 2014 Slide 15 Summary The calibration procedures contributing to an integrated system level simulation are described Some preliminary results on Instantaneous SINR, PHY SLS and Integrated SLS are shown. For each box in the calibration procedures, Companies can choose PHY abstraction method if its accuracy can be proven. The assumptions of average interference seen from all nodes within each OBSS(opt 4) should be one of the metrics to calibrate long term SINR. CDF of Instantaneous SINR per tone could be used to calibrate small scale fading channel assuming a simple MAC such as CCA-only. MAC calibration should focus on MAC mechanism independent of any PHY features. PHY SLS calibration provides a performance baseline of PHY only features. Integrated SLS provides aligned 11ac performance baselines with both PHY and MAC features Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0May 2014 Slide 16 References [1] 11-13/1392r0 Methodology of calibrating system simulation results [2] 11-14/0053r0 Further Considerations on Calibration of System Level Simulation [3] 11-14/0336r0 Calibration of Long-Term SINR for System Simulator [4] 11-13/1001r5 Simulation Scenarios Document Template [5] 11-13/1131r0 PHY abstraction for HEW system level simulation [6] 11-13/1390r0 PHY Abstraction for HEW System Level Simulation [7] 11-13/1059r0 PHY abstraction for HEW evaluation methodology [8] 11-13/1051r1 Evaluation Methodology [9] 11-14/0307r0 PHY Calibration Results [10] 11-13/1359r1 hew-evaluation-methodology [11] 11-14/0335r0 Instantaneous SINR Calibration for System Simulation [12] 11-14/0634r0 Consideration on MAC system calibration Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0 Slide 17 May 2014 Appendix-1 Basic Parameters for Calibration Parameter Value ResidentialEnterpriseIndoorOutdoor Central Frequency2.4 GHz Bandwidth20 MHz Channel Assignment Random assignment of 3 non overlapping channels chan = mod(BSS_index,4). That is, in each office, (ch1, ch2, ch3, ch4) and identical for each office. Ignore adjacent channel interference. Frequency reuse 3Frequency reuse 1 Transmission Power AP / STA23 dBm / 17 dBm24 dBm / 21 dBm17 dBm /15 dBm23 dBm /17 dBm Antenna Configuration Antenna type Antenna gain # of AP/STA antenna omni-directional 0 dB {1, 2 (in Box4&5) } omni-directional 0 dB {1, 2 (in Box4&5) } omni-directional 0 dB {1, 2 (in Box4&5) } omni-directional 0 dB {1, 2 (in Box4&5) } Penetration LossWall / Floor12 dB / 17 dB7 dB / N.A.N.A. / N.A.100% outdoor Noise FigureAP / STA7 dB / 7 dB Channel Model AP-AP AP-STA STA-STA TGn Model B TGn Model D TGn Model B ITU UMi Antenna Height AP STA 1.5m 3m 2m 1.5m 10m 1.5m STA Number10 per apartment4 per cubicle30 per BSS 50 per BSS (Box 1); 10 per BSS (Box 2,4,5) AP-STA min distance (2D)1mN.A.1m10m Association 100% STA associated to the AP in the same apartment 100% STA associate with the strongest AP in the same office 100% of STAs associate with the strongest AP Jiyong Pang, et. al. Huawei

Submission doc.: IEEE /0612r0 Appendix-2 An Example of CCA-only Mechanism Slide 18 May 2014 e.g., N= 15, at one slot, R(1:15) = [ ] AP 11 is CCA idle and STA 12 is selected as receiver STA 8 is CCA busy There has been one transmitter in BSS 1 There has been one transmitter in BSS 3 STA 12 has been selected as receiver There has been one transmitter in BSS 1 There is no transmitter in BSS 2 and STA 7 is CCA idle There is no transmitter in BSS 1 and STA 5 is CCA idle No need to check the rest 6 nodes Jiyong Pang, et. al. Huawei